Deep Learning of Videourodynamics to Classify Bladder Dysfunction Severity in Patients With Spina Bifida

Area of Science:

• Urology
• Medical Imaging
• Artificial Intelligence

Background:

• Videourodynamics is crucial for identifying neurogenic bladder patients at risk of upper tract injury.
• Interpretation of videourodynamics has significant interobserver variability among urologists.
• Developing objective classification methods is essential for consistent patient management.

Purpose of the Study:

• To develop and evaluate deep learning models for categorizing bladder dysfunction severity using videourodynamics data.
• To improve the accuracy and reduce variability in assessing neurogenic bladder severity.
• To aid in the early identification of patients at risk for upper tract damage.

Main Methods:

• A cross-sectional study included 306 videourodynamics studies from pediatric patients with spina bifida.
• Four models were developed: a random forest clinical model, a convolutional neural network for pressure-volume data, a deep learning model for fluoroscopic images, and an ensemble model.
• Bladder dysfunction severity (none/mild, moderate, severe) was defined by expert reviewer consensus, considering factors like compliance and detrusor sphincter dyssynergia.

Main Results:

• The ensemble model, combining pressure-volume and imaging data, achieved 70% accuracy and moderate agreement (weighted kappa 0.54) in classifying bladder dysfunction.
• A clinical model using extracted data performed less effectively, with 61% accuracy and a weighted kappa of 0.37.
• Deep learning models demonstrated potential for automated classification of bladder dysfunction severity.

Conclusions:

• Deep learning models utilizing urodynamic pressure-volume tracings and fluoroscopic images can automatically classify bladder dysfunction with moderately high accuracy.
• These AI-driven approaches offer a promising solution to reduce interobserver variability in videourodynamics interpretation.
• The developed models can assist urologists in more consistent risk stratification for patients with neurogenic bladders.